Whistle Classification in the California Current: A Complete Whistle Classifier for a Large Geographic Region with High Species Diversity.

摘要

The value of passive acoustics as a tool for studying marine mammals relies on the ability to detect and classify sounds associated with these species. Classification efforts typically focus on a single species, or occasionally on a few species found within a geographic region. In an effort to improve our ability to classify species during shipboard population surveys and other passive acoustic monitoring needs, we have developed an automated whistle classification algorithm that includes all whistling species found within a large geographic region, the California Current. In order to train this whistle classifier, single-species encounters within the California Current and near (Eastern Tropical Pacific) were compiled from archived recordings. Whistles were automatically detected using the ‘Whistle and Moan Detector’ (WMD) within the acoustic data processing software platform, PAMGUARD (Gillespie 2008). Measurements from each whistle contour were automatically extracted using the ROCCA (Real-time Odontocete Call Classification Algorithm) module in PAMGUARD. These data were used to train and test a region specific ROCCA whistle classifier to be used on future surveys. This report presents the data used to develop this California Current whistle classifier and the results from testing the classifier on this training data.